I am a Data Scientist & AI Engineer at VNPT, specializing in state-of-the-art Natural Language Processing (NLP), Large Language Models (LLM), and high-performance AI systems.
- 🔭 Specialization: Large Language Models (LLMs), Natural Language Processing (NLP), Speech & Audio AI, Retrieval-Augmented Generation (RAG), and Agentic AI workflows.
- ⚡ Inference & Optimization: High-throughput model serving and low-latency inference using vLLM, TensorRT-LLM, ONNX Runtime, quantization (AWQ, GPTQ, BitsAndBytes), and GPU CUDA acceleration.
- 🧪 Fine-Tuning & Training: Domain adaptation, instruction fine-tuning with PEFT / LoRA / QLoRA, Hugging Face Transformers, and DeepSpeed.
- 🏗️ AI Systems & MLOps: Building scalable LLM microservices and pipelines using FastAPI, Vector Databases (Qdrant, ChromaDB, FAISS), and Docker.
- 📚 Knowledge Hub: This repository acts as my personal engineering handbook — packed with GPU/CUDA troubleshooting, terminal tricks, and Python performance patterns for AI engineers.
| Domain | Core Frameworks & Tools | Specialization & Focus |
|---|---|---|
| 🤖 LLMs & Generative AI | Hugging Face (Transformers, PEFT) • vLLM • LangChain • LlamaIndex • DeepSpeed • Ollama |
SFT, LoRA/QLoRA Fine-Tuning, Prompt Engineering, Agentic Workflows |
| 🧠 Deep Learning & NLP | PyTorch • TensorFlow • spaCy • Scikit-Learn • Python |
Sequence Modeling, Text Classification, NER, Audio & Speech Pipelines |
| ⚡ Acceleration & Inference | NVIDIA CUDA • TensorRT-LLM • ONNX Runtime • BitsAndBytes • Numba |
KV Cache Optimization, AWQ/GPTQ Quantization, Low-Latency Serving |
| 🔍 RAG & Vector Search | Qdrant • ChromaDB • FAISS • Elasticsearch • NumPy • Pandas |
Dense/Sparse Hybrid Retrieval, Semantic Re-ranking, Context Chunking |
| 🚀 AI Serving & MLOps | FastAPI • Triton Inference Server • Docker • MLflow • Weights & Biases |
Asynchronous High-Throughput APIs, Model Versioning, Experiment Tracking |
A collection of practical guides, troubleshooting fixes, and code patterns maintained directly in this repository:
| Category | Guide / Reference | Highlights |
|---|---|---|
| 🐳 Containers & Network | Docker Tutorial | Container lifecycle, Dockerfile tips, GPU container runtime (nvidia-container-runtime), and cache cleanup |
| Git in Docker | Standard Git workflows, cloning, and credential setups inside containerized environments | |
| Ngrok Tunneling | Quick Ubuntu installation, auth token setup, and exposing local services to public URLs | |
| Tmux Multiplexer | Essential session management, split panes, shortcuts, and persistent terminal sessions | |
| ⚡ Linux & GPU Acceleration | NVIDIA & CUDA Troubleshooting | Fixing libcusolver.so symlink errors, TensorFlow GPU device verification, and driver purge/reinstall |
| Console Power-User Tips | Ubuntu port opening (UFW), disk & syslog cleanup, pip cache purge, memory inspection, and audio fixes | |
| GNOME & cuDNN | Dash-to-dock customization commands and cuDNN setup for Ubuntu | |
| 🐍 Python Mastery | Make Python Faster than C++ | Bypassing the Python GIL with Numba JIT compiler into native machine instructions |
| Decorator Design Pattern | Elegant implementation of Python function & class decorators | |
| Protocols & Duck Typing | Modern structural subtyping with Python's typing.Protocol |
|
| Function Arguments Deep Dive | Positional-only, keyword-only, *args, and **kwargs demystified |
|
| IceCream Debugging | Painless, readable variable inspection with ic() instead of print() |
|
| PyPy3 Installation | Setting up PyPy JIT interpreter and optimized packages on Ubuntu | |
| 🐙 Git Standards & Productivity | Git Branch Naming Conventions | Structured standards for dev, test, feature/, bugfix/, and release/ branches |
| 14+ Useful Git Tips | Fast file search (t), blame (b), commit graph aliases, undoing commits, and SSH port 443 timeout override |
|
| 📓 Cloud & Documentation | Google Colab Guide | Google Drive mounting, GPU acceleration check, and notebook productivity |
| Markdown & README Syntax | Syntax cheatsheet for crafting clean GitHub documentation |
Crafted with passion by Nguyen Binh Minh. Always open to discussing AI, NLP, and system engineering!
